Forschungszentrum Jülich

Shaping change: open source for Big Science

Technologies
python, mpi, high performance computing, pytorch, gpu
Topics
neuroscience, data analytics, data-intensive science, earth-system monitoring, space science
Shaping change: open source for Big Science
Conducting research for a changing society: This is what drives us at Forschungszentrum Jülich. As a member of the Helmholtz Association, we aim to tackle the grand societal challenges of our time and conduct research into the possibilities of a digitized society, a climate-friendly energy system, and a resource-efficient economy. Supercomputers help us find solutions to major scientific challenges and are indispensable for modern research. The Jülich Supercomputing Centre (JSC) operates the most powerful computer systems for scientific and technical applications in Europe and makes it available for research purposes to scientists at Forschungszentrum Jülich, in Germany, and throughout Europe. JSC performs various research and development tasks, frequently in tight cooperation with national and international partners. With our open-source Python library Heat, we make it possible for researchers to transition seamlessly from scipy-based, single-node data analysis tools, to memory-distributed, hardware-accelerated, high-performance data analytics.
2022 Program

Successful Projects

Contributor
Neo Sun Han
Mentor
Michael Tarnawa
Organization
Forschungszentrum Jülich
Python array API standard compatibility
Heat is a Python library for high-performance data analytics. It gives users access to multi-node processing and GPU support by seamlessly replacing...
Contributor
Pratham Shah
Mentor
Claudia Comito, Hans
Organization
Forschungszentrum Jülich
RFI Mitigation
This project explores the use of image analysis in a two-dimensional spectrum to detect RFI (Radio Frequency Interference) in large amounts of...
Contributor
Tewodros Mesfin
Mentor
Björn Hagemeier, Juan Pedro Ghm
Organization
Forschungszentrum Jülich
Interactive benchmarking analysis via Ginko Performance Explorer
HeAT, an array-based numerical programming framework for large-scale parallel processing with an easy-to-use NumPy-like API. HeAT utilizes PyTorch as...
Contributor
V. Sai Suraj
Mentor
Daniel Coquelin
Organization
Forschungszentrum Jülich
Memory-Distributed Singular value decomposition.
The major goal of the project is to develop a distributed SVD algorithm that is both efficient and numerically stable in Heat. This will be a major...